Challenge: Recent work has shown that infusing layout features into language models improves processing of visually-rich documents such as scientific papers.
Approach: They propose a method to evaluate layout-infused language models that incorporate layout features into their models to emulate layout distribution shifts.
Outcome: The proposed model performs better under layout distribution shifts than in-distribution conditions.

Similar Papers

VILA: Improving Structured Content Extraction from Scientific PDFs Using Visual Layout Groups (2022.tacl-1)

Copied to clipboard

Challenge: Recent work has improved extraction accuracy by incorporating elementary layout information, for example, each token’s 2D position on the page, into language model pretraining.
Approach: They propose a method that explicitly models VIsual LAyout (VILA) groups, that is, text lines or text blocks, to further improve extraction accuracy.
Outcome: The proposed methods show that inserting special tokens denoting layout group boundaries can lead to a 1.9% Macro F1 improvement in token classification.
Distribution Shifts Are Bottlenecks: Extensive Evaluation for Grounding Language Models to Knowledge Bases (2024.eacl-srw)

Copied to clipboard

Challenge: Existing benchmarks fail to reflect robustness challenges and fairly evaluate models.
Approach: They propose to ground language models to knowledge bases to investigate distribution shifts in language and linguistic aspects of distribution shift.
Outcome: The proposed method fails to evaluate language models in large and small datasets . the proposed model fails to cope with unseen schemas and language variations .
Text Classification Under Class Distribution Shift: A Survey (2026.eacl-long)

Copied to clipboard

Challenge: ML models assume that training and test data are sampled from the same distribution, but in daily practice, this assumption is often broken.
Approach: They survey articles studying open-set text classification to understand the distribution shifts and mitigation approaches for each problem setup.
Outcome: The proposed methods can solve problems caused by the shifting class distribution in open-set text classification and related tasks.
Problem Solved? Information Extraction Design Space for Layout-Rich Documents using LLMs (2025.findings-emnlp)

Copied to clipboard

Challenge: a new open-source layout-aware IE test suite is available for download at https://github.com/gayecolakoglu/layIE-LLM.
Approach: They propose an open-source layout-aware IE test suite that provides a layout-based IE pipeline.
Outcome: The proposed method achieves 13.3–37.5 F1 points more than a baseline configuration using the same LLM.
From Distributional to Overton Pluralism: Investigating Large Language Model Alignment (2025.naacl-long)

Copied to clipboard

Challenge: a large language model's (LLM) output distribution is changed by an alignment process . a recent study shows that aligned models surface information that cannot be recovered from base models without fine-tuning.
Approach: They analyze two aspects of the alignment process that change output distributions . they find alignment suppresses irrelevant and unhelpful content .
Outcome: The proposed model can be imitated without fine-tuning by using in-context examples and lower-resolution semantic hints about response content.
DRPruning: Efficient Large Language Model Pruning through Distributionally Robust Optimization (2025.acl-long)

Copied to clipboard

Challenge: Structured pruning reduces model size but often causes uneven degradation across domains, leading to biased performance.
Approach: They propose a method that dynamically adjusts the data distribution during training to restore balanced performance across heterogeneous and multi-tasking data.
Outcome: Experiments in monolingual and multilingual settings show that the proposed method surpasses similarly sized models in pruning and continued pretraining over perplexity, downstream tasks, and instruction tuning.
LayoutLLM: Large Language Model Instruction Tuning for Visually Rich Document Understanding (2024.lrec-main)

Copied to clipboard

Challenge: Existing methods to enhance document comprehension require fine-tuning for each task and dataset, and are expensive to train and operate.
Approach: They propose a more flexible document analysis method that integrates visual-rich document understanding with large-scale language models (LLMs) by leveraging existing research in document image understanding and LLMs’ superior language understanding capabilities, the proposed model performs an understanding of document images in a single model.
Outcome: The proposed model improves on the baseline model in document image understanding tasks.
Generative Data Augmentation using LLMs improves Distributional Robustness in Question Answering (2024.eacl-srw)

Copied to clipboard

Challenge: Existing domain adaptation methods do not account for unseen natural distribution shifts.
Approach: They perform experiments on 4 different datasets under varying amounts of distribution shift . they analyze how "in-the-wild" generation can help achieve domain generalization .
Outcome: The proposed approach augments reading comprehension datasets with generated data to improve robustness towards natural distribution shifts.
Measuring Distribution Shift in User Prompts and Its Effects on LLM Performance (2026.acl-long)

Copied to clipboard

Challenge: a large-scale evaluation of deployed LLMs under natural prompt distribution shift is needed . natural prompt behavior shifts can cause performance degradation in dynamic, real-world settings .
Approach: They propose a data-centric framework for measuring natural prompt distribution shift . they train models on 4.68M training prompts and evaluate on 57.6k prompts .
Outcome: The proposed framework evaluates natural prompt distribution shift in LLMs over time and between user groups.
Language Models Resist Alignment: Evidence From Data Compression (2025.acl-long)

Copied to clipboard

Challenge: Large language models (LLMs) may exhibit undesirable behaviors due to the inevitable biases and harmful content present in training.
Approach: They propose to investigate the elasticity of large language models by examining their performance.
Outcome: The proposed model performance declines rapidly before reverting to the pre-training distribution, the authors show . the proposed model weight and code are available at pku-lm-res ist-alignment.github.io.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations